The Irf3 transgenic mice (stock number RBRC00858, B6;129S6-Bcl2l12/Irf3 < tm2Ttg > /TtgRbrc) were kindly provided by Prof. Zou Qiang in Chengdu Medical College, which were purchased from RIKEN institute, Japan. The full characterization of these mice has been described elsewhere (Prantner et al., 2010, 2011). All mice used in this study were backcrossed (ten generations) to the C57BL/6J strain (Jackson Laboratories). All studies were performed using male mice. To identify the genotype of mice, PCR, real-time RT-PCR and Western blot were used to detect the genome, mRNA and protein level of IRF3. The young adult (YA) mice used in the study are among 3–4 months, the old mice are 22–24 months, were bred in our vivarium. The Institutional Animal Care and Use Committee (IACUC) of the Army Medical University approved all studies using animals.

The abdominal cavity of each mouse was injected with chloral hydrate (430 mg/kg) for anesthesia. The pupils were scattered by a drop of 1% tropicamide. OCT image acquisition and analysis obtained by Heidelberg Spectralis and Bioptigen Envisu system, averaged transverse scan images were used as a measure of retinal thickness (Zhou et al., 2018). Measurements were applied on four areas of each eye (0.5 mm from optic disk at vertical and horizontal directions), and the average of the four measurements is considered to be the retinal thickness of eye. The total retinal thickness was calculated from the distance between internal limiting membrane (ILM) and Bruch’s membrane (BM). Data are presented as mean ± SEM.

Briefly, mice were killed with overdose CO2, eyes were embedded in Tissue Freezing Medium (Sakura Tissue-Tek O.C.T) at 4°C for 10 min, -80°C for 30 min, and sectioned at 10 μm in a cryostat. Sections were fixed with 4% paraformaldehyde (PFA) for 15 min, washed with PBS. HE staining (Solarbio) was used to observe the structure of the retina. Sections were examined with a LEICA DM1000 microscope. 10× and 40× objective was used for histological measurements. The images were captured by LAS V48 system and were then further processed with Image J. Twelve tissue blocks from six mice (two blocks per mouse) were taken from each group. Three areas of the retina (central, media, periphery) were included in the measurements (2∼3 measurements per tissue bloc). The following parameters were measured: (i) The thickness of retinal cross-section from outer limiting membrane to the inner limiting membrane (OLM-ILM); (ii) the width of each layer of retina.

Animals were adapted to darkness overnight (>12 h) in advance and all operations were performed in dim red light. Each mouse was injected with a mixture of ketamine (60 mg/kg) and xylazine (9 mg/kg) in the abdominal cavity. The body temperature was maintained at 37°C by a warm heating pad. Mydriasis was achieved by a drop of 1% tropicamide. Flash-induced ERG of both eyes was recorded simultaneously using RetiMINER 3.0 system (IRC, China). Stimulus flashes were produced with flash stimulator (Light Burst). Stimulus intensities increased from -2.0 to 1.0 log cd^∗s/m2 in 0.5-log unit steps were used in the dark. A 1-min interval was presented between continuous light stimuli.

Mice were anesthetized as mentioned earlier, and then eyes were fixed with 4% paraformaldehyde (PFA) for 30 mins, the lens and cornea were removed, and the optic cup was post fixed with 4% PFA for 2 h, followed by 30% sucrose dehydration at 4°C overnight. Optic cup was embedded in Tissue Freezing Medium (Sakura Tissue-Tek O.C.T), sectioned at 10 μm in a cryostat. Sections were incubated with blocking solution for 1 h at room temperature. The blocking solution is made of 1.5% BSA and 3% donkey serum, with 0.3% Triton X-100 in PBS. Primary antibodies were diluted in blocking solution for overnight incubation at 4°C, after 3 times rinse in phosphate buffer saline (PBS), sections with fluorophore-labeled secondary antibodies (Thermo scientific) were incubated at room temperature for 2 h. And the nucleus was stained with DAPI for 5 min. Retinas from 4–6 mice were analyzed in each experimental group. The primary antibodies applied in immunofluorescence were: Bassoon (1:400, Enzo Life Sciences), IRF3 (1:400, Abcam), calbindin (1:1000, Abcam), PKCα (1:500, anti-mouse, Abcam), PSD-95 (1:400, Millipore), P-H2A.X (1:1000, Cell Signaling Technology), syntaxin I (1:500, anti-mouse, Abcam), Sox9 (1:500 Millipore). The sections after the staining were mounted with DPX (Sigma) or Vectashield (Vector Laboratories). Sections were captured by the confocal microscope (Leica SP8) and the corresponding images were collected by Leica Application Suite X. Series of images were overlaid to a z-projection with maximum intensities. The analysis of retinal micrographs was taken with ImageJ and the data was analyzed by GraphPad Prism 6.0.

Mice were sacrificed by inhalation of CO2, and eyes were removed to obtain the whole retina. Retinas (n = 6 in each group) were ground in RIPA lysis buffer (Thermo scientific) with 1 mM PMSF and 2 mM EDTA. The lysate was centrifuged at 12,000 rpm for 15 min at 4°C, and the supernatant was collected. Protein in per sample was 2.5–2.7 mg added with 5× loading buffer for 8 min boiling, loaded on 4–12% SDS-polyacrylamide gels (Beyotime), transferred onto PVDF membranes. The membranes were blocked with 5% non-fat dry milk dissolved in Tris-buffered saline for 2 h at room temperature. Then the membranes were sequentially probed overnight at 4°C with primary antibodies diluted with blocking solution. The primary antibodies were arranged for the following molecules: IRF3 (1:1000, Abcam), P53 (1:1000, Santa Cruz Biotechnology), P16^INK4a (1:1000, Proteintech), PML (1:1000, Santa Cruz Biotechnology), P-H2A.X (1:1000, Cell Signaling Technology), IREα (1:1000, Cell Signaling Technology), GAPDH (1:1,000, Invitrogen). Membranes were washed with TBS-T (pH = 7.5) containing 0.2% Tween for three times, and incubated with appropriate secondary antibody for 90 min at room temperature. The band intensities were measured using exposure machine (Omega Lum^TM G). Quantification of band intensity was analyzed by using ImageJ and GraphPad Prism 6.0.

Electron microscopy was performed as previously described (Chen et al., 2015). Mice were killed by an overdose of CO2. The retinas of each group of mice were removed quickly and fixed with fresh fixative (2% glutaraldehyde in 0.1 M cacodylate buffer) overnight at 4°C. Retinas were rinsed in phosphate buffer saline (PBS), post fixed in 1% OsO4 in PBS for 2 h at room temperature, co-stained with uranyl acetate, dehydrated strictly in a graded ethanol series, fully infiltrated with propylene oxide, and slowly embedded in epon. Thin sections were stained with toluidine blue. Ultrathin sections (∼70 nm) were obtained by an ultramicrotome (LKB-V, LKB Produkter AB, Bromma) and were viewed with a transmission electron microscope (TECNAI10, Philips). Sixteen sections from each mouse retina were analyzed and four mice were selected for each group. The position of the OPL layer was observed in each section under 12,000 magnification.

All analyses were carried out with Prism software version 6 (GraphPad). For two-sample comparisons (Western blot results, Figure 1B), an unpaired, two-tailed Student’s t-test was applied. Data in other figures was analyzed with two-way ANOVA followed by either a Bonferroni or Kruskal-Wallis post hoc test. All data are presented as mean ±SEM. P-values of less than 0.05 (P < 0.05) were considered to be statistically significant.

In order to explore the role of IRF3 in retinas, we evaluated the total change of IRF3 protein by using western blotting from a general perspective. Interestingly, IRF3 was significantly reduced in old mouse retinas compared with those of young mice (Figure 1A,B). To further ascertain the role that IRF3 plays in retinas, we uncovered the location of IRF3 in YA WT mice by staining IRF3 with DAPI. The results indicated that IRF3 was expressed in outer plexiform layer (OPL) (Figure 1C). Moreover, consistent with the western blotting results, we noticed that the expression of IRF3 in the retina of Old WT mice had decreased significantly and that some had entered the ONL by comparing retinal fluorescence staining for YA and Old WT mice (Figure 1D). Furthermore, we found that IRF3 may be mainly localized at dendritic terminals of rod bipolar cells by labeling with PKCα (Figure 1E). Taken together, the location of IRF3 expression in OPL and the declining IRF3 expression in aged retina suggest that IRF3 may play an important role in retinal function.

To determine the role that IRF3 plays in retinal function, we examined the retinas of the young adult (YA) and Old Irf3^-/- mice and their corresponding control littermates, YA WT and Old WT, via ERG. After flash ERG stimuli, the responses of photoreceptor cells (a-wave) and summed synaptic responses for both inner and outer retina (b-wave) (Samuel et al., 2014) were significantly weakened in Irf3^-/- mice compared with WT mice of the same age, represented by the lower amplitude of a-wave and b-wave in both YA and Old Irf3^-/- groups (Figure 2A–C). Interestingly, with increased ERG stimulation intensity, the a-wave and b-wave amplitudes also increased significantly, but the responses in Irf3^-/- mice were consistently weaker than in WT mice of the same age (Figure 2A–C). As ERG recordings were tested under dark-adapted conditions, the results indicated that the Irf3^-/- mice largely lacked rod-driven responses. To be noted, PNA (peanut agglutinin) and cone arrestin staining revealed that the deletion of IRF3 did not cause cone cells loss, which instead resulted from age (Supplementary Figures S1A–C).

These results demonstrate that YA IRF3-deficient mice exhibit alterations in retinal physical function similar to those that have been reported for old WT mice (Jackson and Owsley, 2003; Owsley, 2011). In addition, Old Irf3^-/- mice presented worse retinal physical function than old WT mice, indicating that IRF3 is crucial for retinal function. However, whether the structure of the retina is affected in Irf3^-/- mice requires further study.

To investigate the overall effect of IRF3 deletion on retinal structure, retinal thickness in both the live and deceased mice was evaluated. As the OCT data showed, the thickness of the retina was significantly reduced in Irf3^-/- mice compared with WT mice of the same age (Figure 3A–C). To further understand which layer was responsible for the thickness change, we used HE staining. In agreement with the OCT result, the total thickness of three different retinal regions (central, media, periphery) was reduced in Irf3 mutants (Figure 3D,E). Interestingly, regardless of the area in the retina, the thickness of the OPL and the outer nuclear layer (ONL) were both reduced in Irf3^-/- mice, and the OPL was remarkably reduced (Figure 3E). However, a reduction in retinal thickness did not appear in Irf3^-/- mice at the beginning or the end of the developmental window for photoreceptor synaptogenesis (P7, P21) (Supplementary Figures S2A,B). These results suggest that IRF3 is essential for the structure of YA and Old retinas, especially the OPL.

Based on the aforementioned structural and functional alteration, we sought to uncover the associated underlying pathological changes. As the thickness of the OPL is most clearly altered in Irf3^-/- retinas and photoreceptor cells and interneurons form synapses exclusively in the OPL (Hoon et al., 2014; Samuel et al., 2014), we investigated synaptic structure in the OPL by immunofluorescence staining. As shown by Bassoon and DAPI co-labeling, photoreceptor synaptic terminals (pre-synaptic) obviously withdraw into the ONL of the retina in Irf3^-/- mice compared with WT mice at both YA and Old stages (Figure 4A,B). Moreover, to identify synaptic contacts between photoreceptor cells (pre-synaptic) and dendritic terminals (post-synaptic) of rod bipolar cells in the OPL, we used double labeling with bassoon and PKCα. Both synaptic markers revealed that sprouts in YA and Old IRF3-deficient mice were dotted with numerous ectopic synapses, consistent with the location of Bassoon (Figure 4A). Similarly, the PKCα-positive dendrites of rod bipolar cells and calbindin-positive neurites of horizontal cells (the post-synaptic terminals) both extended further into the ONL in Irf3^-/- mice compared with WT mice (Figure 4A). Unexpectedly, Similar synaptic changes were not seen in Irf3^-/- mice at P7 and P21 (Supplementary Figure S2C).

In addition to the location alternation of the OPL synapse, we also found a reduced expression of PSD-95, a protein expressed on photoreceptor synaptic terminals (Koulen et al., 1998), in the OPL of retinas from the Irf3^-/- mice compared with WT mice of the same age (Figure 4C,D). However, the PSD-95 level was not significantly reduced in the outer retina of each group (Figure 4E–G), indicating that the decrease in synaptic density in the OPL was not caused by the change of the total number of synapses in the outer retina. To illustrate the ultrastructure of the synapses in the OPL, electron microscopy was performed. Triad structure and ribbon, which is typical of the morphology of rod spherule synapses in the OPL, was observed in all groups (Figure 4H, red arrow). Interestingly, the density of the triad structure was also remarkably reduced in Irf3^-/- mice compared with WT mice of the same age, which indicates that IRF3 may be crucial for maintaining synapse density in the OPL (Figure 4H,I).

Based on our results, IRF3 deletion results in the alternation of the normal architecture and density of synapse in the OPL. Collectively, the aged-related morphological and functional changes in Irf3^-/- mice are more obvious compared with WT mice (both the YA and Old groups), indicating IRF3 deficiency may play a role in accelerating retinal senescence.

Our results reveal a potential relationship between IRF3 and senescence. To directly address this hypothesis, we took advantage of a series of senescence markers which classically reflect the stages of cellular senescence in retina. The protein level of p53, p16^INK4a, inositol-requiring enzyme α (IREα), p-H2A.X and promyelocytic leukemia protein (PML) were evaluated in WT and Irf3^-/- retinas (Campisi and d’Adda di Fagagna, 2007; Oubaha et al., 2016; Yang et al., 2016). Qualitative and quantitative analysis indicated that IRF3 deletion in the retina lead to higher expressions of these five markers than WT mice of the same age. Hardly any IRF3 expression was detected in Irf3^-/- mice (Figure 5A,B and Supplementary Figure S3). In addition, dramatic reduced IRF3 expression was found in Old WT mice compared with YA WT mice (Figure 5A,B, IRF3 panel and Supplementary Figure S3 , which implied that IRF3 expression is affected during normal aging.

To further determine which neuronal cells of the retina might be involved in senescence, we applied immunofluorescence staining to p-H2A.X, a phosphorylated histone that appears in the nucleus of senescence cell (Fang et al., 2012; Yoshizawa et al., 2012). Our result revealed that p-H2A.X expressions were largely increased in Irf3^-/- mice compared with WT mice at both YA and Old stages (Figure 5C,D). Meanwhile, in Irf3^-/- retinas, p-H2A.X was mainly located in the INL (Figure 5C), indicating that some neuron cells of INL might be affected. Interestingly, by further co-localization of p-H2A.X with four types of cells in the INL of Irf3^-/- mice, we found that most of the p-H2A.X positive cells co-localized syntaxin I in the amacrine cells (Supplementary Figure S4). Together, these results indicate that the deletion of IRF3 exacerbates the senescence of retina in YA and Old mice.